xref: /freebsd/sys/kern/subr_param.c (revision e0c4386e7e71d93b0edc0c8fa156263fc4a8b0b6)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 #include <sys/cdefs.h>
38 #include "opt_param.h"
39 #include "opt_msgbuf.h"
40 #include "opt_maxphys.h"
41 #include "opt_maxusers.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/buf.h>
46 #include <sys/kernel.h>
47 #include <sys/limits.h>
48 #include <sys/msgbuf.h>
49 #include <sys/sysctl.h>
50 #include <sys/proc.h>
51 #include <sys/vnode.h>
52 
53 #include <vm/vm.h>
54 #include <vm/vm_param.h>
55 #include <vm/pmap.h>
56 
57 /*
58  * System parameter formulae.
59  */
60 
61 #ifndef HZ
62 #  define	HZ 1000
63 #  ifndef HZ_VM
64 #    define	HZ_VM 100
65 #  endif
66 #else
67 #  ifndef HZ_VM
68 #    define	HZ_VM HZ
69 #  endif
70 #endif
71 #define	NPROC (20 + 16 * maxusers)
72 #ifndef NBUF
73 #define NBUF 0
74 #endif
75 #ifndef MAXFILES
76 #define	MAXFILES (40 + 32 * maxusers)
77 #endif
78 
79 static int sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS);
80 
81 int	hz;				/* system clock's frequency */
82 int	tick;				/* usec per tick (1000000 / hz) */
83 time_t	tick_seconds_max;		/* max hz * seconds an integer can hold */
84 struct bintime tick_bt;			/* bintime per tick (1s / hz) */
85 sbintime_t tick_sbt;
86 int	maxusers;			/* base tunable */
87 int	maxproc;			/* maximum # of processes */
88 int	maxprocperuid;			/* max # of procs per user */
89 int	maxfiles;			/* sys. wide open files limit */
90 int	maxfilesperproc;		/* per-proc open files limit */
91 int	msgbufsize;			/* size of kernel message buffer */
92 int	nbuf;				/* number of bcache bufs */
93 int	bio_transient_maxcnt;
94 int	ngroups_max;			/* max # groups per process */
95 int	nswbuf;
96 pid_t	pid_max = PID_MAX;
97 u_long	maxswzone;			/* max swmeta KVA storage */
98 u_long	maxbcache;			/* max buffer cache KVA storage */
99 u_long	maxpipekva;			/* Limit on pipe KVA */
100 u_long	maxphys;			/* max raw I/O transfer size */
101 int	vm_guest = VM_GUEST_NO;		/* Running as virtual machine guest? */
102 u_long	maxtsiz;			/* max text size */
103 u_long	dfldsiz;			/* initial data size limit */
104 u_long	maxdsiz;			/* max data size */
105 u_long	dflssiz;			/* initial stack size limit */
106 u_long	maxssiz;			/* max stack size */
107 u_long	sgrowsiz;			/* amount to grow stack */
108 
109 SYSCTL_INT(_kern, OID_AUTO, hz, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &hz, 0,
110     "Number of clock ticks per second");
111 SYSCTL_INT(_kern, OID_AUTO, hz_max, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, HZ_MAXIMUM,
112     "Maximum hz value supported");
113 SYSCTL_INT(_kern, OID_AUTO, hz_min, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, HZ_MINIMUM,
114     "Minimum hz value supported");
115 SYSCTL_INT(_kern, OID_AUTO, nbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nbuf, 0,
116     "Number of buffers in the buffer cache");
117 SYSCTL_INT(_kern, OID_AUTO, nswbuf, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &nswbuf, 0,
118     "Number of swap buffers");
119 SYSCTL_INT(_kern, OID_AUTO, msgbufsize, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &msgbufsize, 0,
120     "Size of the kernel message buffer");
121 SYSCTL_LONG(_kern, OID_AUTO, maxswzone, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxswzone, 0,
122     "Maximum memory for swap metadata");
123 SYSCTL_LONG(_kern, OID_AUTO, maxbcache, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxbcache, 0,
124     "Maximum value of vfs.maxbufspace");
125 SYSCTL_INT(_kern, OID_AUTO, bio_transient_maxcnt, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
126     &bio_transient_maxcnt, 0,
127     "Maximum number of transient BIOs mappings");
128 SYSCTL_ULONG(_kern, OID_AUTO, maxtsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxtsiz, 0,
129     "Maximum text size");
130 SYSCTL_ULONG(_kern, OID_AUTO, dfldsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dfldsiz, 0,
131     "Initial data size limit");
132 SYSCTL_ULONG(_kern, OID_AUTO, maxdsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxdsiz, 0,
133     "Maximum data size");
134 SYSCTL_ULONG(_kern, OID_AUTO, dflssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &dflssiz, 0,
135     "Initial stack size limit");
136 SYSCTL_ULONG(_kern, OID_AUTO, maxssiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &maxssiz, 0,
137     "Maximum stack size");
138 SYSCTL_ULONG(_kern, OID_AUTO, sgrowsiz, CTLFLAG_RWTUN | CTLFLAG_NOFETCH, &sgrowsiz, 0,
139     "Amount to grow stack on a stack fault");
140 SYSCTL_PROC(_kern, OID_AUTO, vm_guest,
141     CTLFLAG_RD | CTLTYPE_STRING | CTLFLAG_MPSAFE, NULL, 0,
142     sysctl_kern_vm_guest, "A",
143     "Virtual machine guest detected?");
144 
145 /*
146  * The elements of this array are ordered based upon the values of the
147  * corresponding enum VM_GUEST members.
148  */
149 static const char *const vm_guest_sysctl_names[] = {
150 	[VM_GUEST_NO] = "none",
151 	[VM_GUEST_VM] = "generic",
152 	[VM_GUEST_XEN] = "xen",
153 	[VM_GUEST_HV] = "hv",
154 	[VM_GUEST_VMWARE] = "vmware",
155 	[VM_GUEST_KVM] = "kvm",
156 	[VM_GUEST_BHYVE] = "bhyve",
157 	[VM_GUEST_VBOX] = "vbox",
158 	[VM_GUEST_PARALLELS] = "parallels",
159 };
160 _Static_assert(nitems(vm_guest_sysctl_names) == VM_LAST,
161     "new vm guest type not added to vm_guest_sysctl_names");
162 
163 /*
164  * Boot time overrides that are not scaled against main memory
165  */
166 void
167 init_param1(void)
168 {
169 
170 	TSENTER();
171 
172 	/*
173 	 * arm64 and riscv currently hard-code the thread0 kstack size
174 	 * to KSTACK_PAGES, ignoring the tunable.
175 	 */
176 	TUNABLE_INT_FETCH("kern.kstack_pages", &kstack_pages);
177 
178 	hz = -1;
179 	TUNABLE_INT_FETCH("kern.hz", &hz);
180 	if (hz == -1)
181 		hz = vm_guest > VM_GUEST_NO ? HZ_VM : HZ;
182 
183 	/* range check the "hz" value */
184 	if (__predict_false(hz < HZ_MINIMUM))
185 		hz = HZ_MINIMUM;
186 	else if (__predict_false(hz > HZ_MAXIMUM))
187 		hz = HZ_MAXIMUM;
188 
189 	tick = 1000000 / hz;
190 	tick_sbt = SBT_1S / hz;
191 	tick_bt = sbttobt(tick_sbt);
192 	tick_seconds_max = INT_MAX / hz;
193 
194 	/*
195 	 * Arrange for ticks to wrap 10 minutes after boot to help catch
196 	 * sign problems sooner.
197 	 */
198 	ticks = INT_MAX - (hz * 10 * 60);
199 
200 	vn_lock_pair_pause_max = hz / 100;
201 	if (vn_lock_pair_pause_max == 0)
202 		vn_lock_pair_pause_max = 1;
203 
204 #ifdef VM_SWZONE_SIZE_MAX
205 	maxswzone = VM_SWZONE_SIZE_MAX;
206 #endif
207 	TUNABLE_LONG_FETCH("kern.maxswzone", &maxswzone);
208 #ifdef VM_BCACHE_SIZE_MAX
209 	maxbcache = VM_BCACHE_SIZE_MAX;
210 #endif
211 	TUNABLE_LONG_FETCH("kern.maxbcache", &maxbcache);
212 	msgbufsize = MSGBUF_SIZE;
213 	TUNABLE_INT_FETCH("kern.msgbufsize", &msgbufsize);
214 
215 	maxtsiz = MAXTSIZ;
216 	TUNABLE_ULONG_FETCH("kern.maxtsiz", &maxtsiz);
217 	dfldsiz = DFLDSIZ;
218 	TUNABLE_ULONG_FETCH("kern.dfldsiz", &dfldsiz);
219 	maxdsiz = MAXDSIZ;
220 	TUNABLE_ULONG_FETCH("kern.maxdsiz", &maxdsiz);
221 	dflssiz = DFLSSIZ;
222 	TUNABLE_ULONG_FETCH("kern.dflssiz", &dflssiz);
223 	maxssiz = MAXSSIZ;
224 	TUNABLE_ULONG_FETCH("kern.maxssiz", &maxssiz);
225 	sgrowsiz = SGROWSIZ;
226 	TUNABLE_ULONG_FETCH("kern.sgrowsiz", &sgrowsiz);
227 
228 	/*
229 	 * Let the administrator set {NGROUPS_MAX}, but disallow values
230 	 * less than NGROUPS_MAX which would violate POSIX.1-2008 or
231 	 * greater than INT_MAX-1 which would result in overflow.
232 	 */
233 	ngroups_max = NGROUPS_MAX;
234 	TUNABLE_INT_FETCH("kern.ngroups", &ngroups_max);
235 	if (ngroups_max < NGROUPS_MAX)
236 		ngroups_max = NGROUPS_MAX;
237 
238 	/*
239 	 * Only allow to lower the maximal pid.
240 	 * Prevent setting up a non-bootable system if pid_max is too low.
241 	 */
242 	TUNABLE_INT_FETCH("kern.pid_max", &pid_max);
243 	if (pid_max > PID_MAX)
244 		pid_max = PID_MAX;
245 	else if (pid_max < 300)
246 		pid_max = 300;
247 
248 	TUNABLE_INT_FETCH("vfs.unmapped_buf_allowed", &unmapped_buf_allowed);
249 	TSEXIT();
250 }
251 
252 /*
253  * Boot time overrides that are scaled against main memory
254  */
255 void
256 init_param2(long physpages)
257 {
258 
259 	TSENTER();
260 	/* Base parameters */
261 	maxusers = MAXUSERS;
262 	TUNABLE_INT_FETCH("kern.maxusers", &maxusers);
263 	if (maxusers == 0) {
264 		maxusers = physpages / (2 * 1024 * 1024 / PAGE_SIZE);
265 		if (maxusers < 32)
266 			maxusers = 32;
267 #ifdef VM_MAX_AUTOTUNE_MAXUSERS
268                 if (maxusers > VM_MAX_AUTOTUNE_MAXUSERS)
269                         maxusers = VM_MAX_AUTOTUNE_MAXUSERS;
270 #endif
271                 /*
272                  * Scales down the function in which maxusers grows once
273                  * we hit 384.
274                  */
275                 if (maxusers > 384)
276                         maxusers = 384 + ((maxusers - 384) / 8);
277         }
278 
279 	/*
280 	 * The following can be overridden after boot via sysctl.  Note:
281 	 * unless overridden, these macros are ultimately based on maxusers.
282 	 * Limit maxproc so that kmap entries cannot be exhausted by
283 	 * processes.
284 	 */
285 	maxproc = NPROC;
286 	TUNABLE_INT_FETCH("kern.maxproc", &maxproc);
287 	if (maxproc > (physpages / 12))
288 		maxproc = physpages / 12;
289 	if (maxproc > pid_max)
290 		maxproc = pid_max;
291 	maxprocperuid = (maxproc * 9) / 10;
292 
293 	/*
294 	 * The default limit for maxfiles is 1/12 of the number of
295 	 * physical page but not less than 16 times maxusers.
296 	 * At most it can be 1/6 the number of physical pages.
297 	 */
298 	maxfiles = imax(MAXFILES, physpages / 8);
299 	TUNABLE_INT_FETCH("kern.maxfiles", &maxfiles);
300 	if (maxfiles > (physpages / 4))
301 		maxfiles = physpages / 4;
302 	maxfilesperproc = (maxfiles / 10) * 9;
303 	TUNABLE_INT_FETCH("kern.maxfilesperproc", &maxfilesperproc);
304 
305 	/*
306 	 * Cannot be changed after boot.
307 	 */
308 	nbuf = NBUF;
309 	TUNABLE_INT_FETCH("kern.nbuf", &nbuf);
310 	TUNABLE_INT_FETCH("kern.bio_transient_maxcnt", &bio_transient_maxcnt);
311 	maxphys = MAXPHYS;
312 	TUNABLE_ULONG_FETCH("kern.maxphys", &maxphys);
313 	if (maxphys == 0) {
314 		maxphys = MAXPHYS;
315 	} else if (__bitcountl(maxphys) != 1) {	/* power of two */
316 		if (flsl(maxphys) == NBBY * sizeof(maxphys))
317 			maxphys = MAXPHYS;
318 		else
319 			maxphys = 1UL << flsl(maxphys);
320 	}
321 	if (maxphys < PAGE_SIZE)
322 		maxphys = MAXPHYS;
323 
324 	/*
325 	 * Physical buffers are pre-allocated buffers (struct buf) that
326 	 * are used as temporary holders for I/O, such as paging I/O.
327 	 */
328 	TUNABLE_INT_FETCH("kern.nswbuf", &nswbuf);
329 
330 	/*
331 	 * The default for maxpipekva is min(1/64 of the kernel address space,
332 	 * max(1/64 of main memory, 512KB)).  See sys_pipe.c for more details.
333 	 */
334 	maxpipekva = ptoa(physpages / 64);
335 	TUNABLE_LONG_FETCH("kern.ipc.maxpipekva", &maxpipekva);
336 	if (maxpipekva < 512 * 1024)
337 		maxpipekva = 512 * 1024;
338 	if (maxpipekva > (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 64)
339 		maxpipekva = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) /
340 		    64;
341 	TSEXIT();
342 }
343 
344 /*
345  * Sysctl stringifying handler for kern.vm_guest.
346  */
347 static int
348 sysctl_kern_vm_guest(SYSCTL_HANDLER_ARGS)
349 {
350 	return (SYSCTL_OUT_STR(req, vm_guest_sysctl_names[vm_guest]));
351 }
352